中国科学技术大学学报 ›› 2019, Vol. 49 ›› Issue (3): 237-243.DOI: 10.3969/j.issn.0253-2778.2019.03.009

• 原创论文 • 上一篇    下一篇

基于多路径优化的视频稳像方法

干文杰   

  1. 中国科学技术大学数学科学学院,安徽合肥 230026
  • 收稿日期:2017-12-21 修回日期:2018-05-02 出版日期:2019-03-30 发布日期:2019-03-30
  • 通讯作者: 张举勇
  • 作者简介:干文杰,男,1992年生,硕士.研究方向: 图像和视频处理. E-mail: ganwj@mail.ustc.edu.cn
  • 基金资助:
    国家重点研发计划(2016YFC0800501),国家自然科学基金(61672481)资助.

A multi-path optimization based video stabilization method

GAN Wenjie   

  1. School of Mathematical Sciences,University of Science and Technology of China, Hefei 230026, China
  • Received:2017-12-21 Revised:2018-05-02 Online:2019-03-30 Published:2019-03-30

摘要: 为了对日常生活中手持设备拍摄的视频进行稳像处理,提出了一种基于多路径优化的稳像方法.首先,进行运动估计,将输入视频的每一帧划分成均匀的网格,通过估计网格顶点的运动矢量得到帧间的运动;接着,进行运动平滑,将上一步的运动累加起来得到相机的运动轨迹,然后利用设计的多路径优化模型进行平滑处理;最后,进行运动补偿,对视频的每一帧进行补偿,得到稳像后的视频.将该方法和最近一些较好的方法进行处理时间和稳像结果的对比实验,结果表明,该方法耗时少,能取得令人满意的稳像效果.

关键词: 视频稳像, 相机运动轨迹, 网格, 多路径

Abstract: A multi-path optimization based video stabilization method was introduced, which is used to stabilize videos shot by hand-held devices in daily life. First, motions between two adjacent frames were estimated. Each frame of the input video was divided into a uniform mesh, and the motion of each mesh was estimated. Then the camera path was smoothed. The camera path was calculated by accumulating the motions estimated in the previous step, and a multi-path optimization based path smoothing model was designed to smooth the camera path. Finally, the stabilized video was calculated by warping each frame of the input video. The proposed method has been compared with some popular methods on running time and stabilization results. The comparison shows that the proposed method is efficient and can produce satisfactory results.

Key words: video stabilization, camera path, mesh, multi-path